I realized library size may be an issue and most CNV tools seem to ignore this.
Say, we try to call CNV out of a tumor tissue those genome is almost doubled. If we directly compare the depth of each bin between tumor and normal control and the library size of tumor (fastq size) and normal tissue is equal, the depth of genomically doubled tumor tissue will still have the same depth as the normal tissue. If CNV caller ignore this library size issue, CNV called from genome doubled tissue will be incorrect (estimated diploid baseline is actually double genome). Can anyone share some insight on this? Thanks